The Application Of Depth Camera-based Image Processing Technology In Passenger Flow Statistics

It is the key to build a high quality public transportation system in a city that the statistics of urban passenger flow can grasp the situation of urban passenger flow in real time,realize intelligent public transportation scheduling and reasonably plan the urban road network.This paper proposes a passenger flow statistics algorithm based on deep video image processing.The passenger flow statistics based on depth video mainly consists of three functional modules: depth video denoising,depth video hole filling,and passenger head area extraction.In the depth video denoising module,according to the causes of noise,this paper divides the depth video noise into internal noise and external noise.For the internal noise,the depth video layered denoising algorithm is proposed.Based on the depth threshold and noise area threshold,the noise area is accurately denoised.For the external noise,the algorithm of eliminating the environment interference noise is proposed.The noise area is divided into several small rectangular elements,and the effective depth value around the noise area is used to fill the depth value of the noise area accurately,so as to eliminate the environmental interference noise.In the depth video cavity filling module,aiming at the problem of depth video cavity caused by direct sunlight,this paper proposes a hole filling method based on super-pixel segmentation.Based on the correspondence between color video and depth video in rgb-d video,a strong correlation super pixel unit is generated by dividing the super pixels of color video.The strong correlation super-pixel unit helps to find the most relevant neighborhood in the depth video hole area,and uses the mean value of neighborhood depth value to accurately fill the hole.In the module of extracting the head area of passengers,Hough transform based on gradient information is used to extract the head area of passengers in the depth video,and the depth threshold segmentation is used to eliminate the error extraction results caused by spherical baggage,and the final extracted head area of passengers is generated.For the extracted head area of passengers,the double counting line passenger flow counting system is used to count the passenger flow.